Demand Forecasting and Material Requirement

Planning Optimization using Open Source Tools

Jorge Ivan Romero-Gelvez , Edisson Alexander Delgado-Sierra, Jorge Aurelio

Herrera-Cuartas, and Olmer Garcia-Bedoya

Universidad de Bogotá Jorge Tadeo Lozano, Bogotá, Colombia

jorgei.romerog@utadeo.edu.co

Abstract. The purpose of this work is to contribute to the extended use

of hybrid models to solve MRP issues dealing with stochastic demand
over main stock-keeping units. The methodology development first apply
SARIMA (Seasonal Autoregressive Integrated Moving Average Model),
Long short term memory networks and Fb-prophet as forecasting meth-
ods to predict the demand for the master production schedule, next
applies an integer programming model with JuMP (Julia Mathematical
programming) for solving the MRP using the Lot for lot approach (L4L).
The main contribution of this work is to show a way to solve dynamic
demand problems over the Forecasting-MIP approach.

Keywords: SARIMA · LSTMN · Fb-prophet · Forecasting · MIP

· JuMP.

1 Introduction

The heart of a Material Requirements planning system is the demand forecasting

for the main stock-keeping-unit. However, it is a complicated, time spending and
inaccurate task. The randomness in the forecast has led to improve them using
a variety of tools, such as the Box-Jenkins methodology [3], machine learning
techniques [2], and even to change the paradigm of their use in Material Require-
ments planning [23]. In forecasting we notice the lot-sizing techniques and their
performance [15] and Box-jenkins implementations for MRP had been used for
MRP by several authors, [5], but the case study had no values for cost maintain
inventory or cost of order. Lee work over eoq lot-sizing calculation errors [16] and
there are several applications in Order-Up-To (OUT) [17],machine learning [22].
The incremental use of computers in the 1960’s industry allows them to
improve complex scheduling and inventory control [12], which leads to Orlicki
and others to develop a manufacture planning approach based on an independent
demand over main items and a dependent demand over the material [21].
The development and adoption of Material requirements planning were slow
but in 1972 the American production and inventory control society APICS starts
to implement this model and initiates the mrp crusade that continues in our
time [18]. The early development of MRP (today also know as “little-mrp” or
Copyright © 2019 for this paper by its authors. Use permitted under Creative Commons License
Attribution 4.0 International (CC BY 4.0)
2019 ICAI Workshops, pp. 94–107, 2019.
 Demand Forecasting and Material Requirement Planning Optimization 95

MRP-I) refers to a model that comes from a production plan and focuses over
get the optimal requirements for every Stock keeping unit involved in the bill of
manufacture (also called BOM), [29]. The MRP improvement years later is called
MRP II, but the letters MRP in MRP II stand for Manufacturing Resources
Planning to make it clear that resources are considered in addition to materials
as in mrp. The word “resource” is used to emphasize that any type of productive
capability can be considered, not just machines. This work uses only the MRP I
model, focus the effort in the forecast of demands and manage large sets of data
for every SKU in the company portfolio of products.

2 Literature review and basic background

This section describes the basic concepts of Material requirement planning, De-
mand Driven Material requirement planning and Forecasting techniques needed
to solve the dynamic demand problem presented in the case study.

2.1 Fundamentals of Material requirement planning

This item first describes every part in a production system and later focuses
on MRP implementation. A production plan describes in detail the quantity of
principal sku (final product to sell) and produced in subsets of sku, the exact
time of production and lot sizing. The production plan can be divided in mas-
ter production schedule (MPS), Material requirement planning (MRP) and the
detailed plan of jobs in the production floor. See Fig. 1.
The APICS dictionary defines MRP as: "A set of techniques that uses a bill
of material data, inventory data, and the master production schedule to calculate
requirements for materials" [4]. MRP requires three basic inputs. First the master
production schedule, second a bill of material (BOM) for each sku (part number)
what other sku are required as direct components, and third the actual level of

Fig. 1. The conventional planning schema. Based on [13]

96 J. Romero-Gelvez et al.

(a) BOM for a Simple Example. (b) The basic MRP record form: [13].

Fig. 2. Example of a bill of manufacture and basic records for feeding the MRP system.

inventory for every sku. See Fig. 2. According to [13] an MRP system serves
a central role in material planning and control. It translates the overall plans
for production into the detailed individual steps necessary to accomplish those
plans. It provides information for developing capacity plans, and it links to the
systems that actually get the production accomplished. [20]

Lot for lot MRP Optimization model The formulation for MRP is based
in the model proposed by [29] and follow the form:

P X
X T
minimize (T − t)Xi,t
x
i=1 t=1
subject to
t−LT (i) t  P 
X X X
Xi,τ + I(i, 0) ≥ D(i, τ ) + R(i, j)Xi,j
τ =1 τ =1 j=1

Xi,t − δi,t LS(i) ≥ 0

Xi,t
δi,t − ≥0
M
δi,t ∈ 0, 1
∀i ∈ 1, ..., P , ∀t ∈ 1, ..., T .

P Number of SKUs
T Number of time buckets (i.e., planing horizon)
LTi Lead time for each SKU i
Ri,j Number i necesarios para hacer un j
Di,t External demand for i over tperiod
Ii, 0 Initial inventory for every SKU i
LSi Minimum lot size for every SKU i
M A large number
 Demand Forecasting and Material Requirement Planning Optimization 97

The planning horizon i calculated by [25] and the overall costs are as well
found to significantly increase with forecast error [1]

2.2 Forecasting traditional methods

The core of MPS is the forecasting job to predict future demands. This work
uses the Box-Jenkins methodology to applying the seasonal ARIMA model in
order to obtain the future demand of principal sku.

Time series analysis According to [13] the time series methods are called
common methods because they no longer require other information of past val-
ues. Time series is a term to refer to the collection of observations of economic
or physical phenomena drawn at discrete points in time. The idea is that past
information can be used to forecast future values of the series. In time series
analysis we try to isolate the patterns that arise most frequently, these include
the following:

– Trend: refers to the trend of a series of time that exhibits a stable pattern
of growth or decrease.
– Seasonality: A seasonality pattern are those that are repeated at fixed inter-
vals.
– Cycles: The variation of cycles is similar to seasonality, except that the du-
ration and the magnitude of the cycle varies. One associates cycles with
economic variations that are also present in seasonal fluctuations.
– Randomness: A random series is where you do not have a recognized pat-
tern of data. One can generate a random series of data that have a specific
structure . The data that seems to have apparently a randomness , actually
have a specific structure. Actually the random data fluctuate around a fixed
average.

Box-Jenkins Methodology This method was given thanks to two known

statisticians [3]. The models proposed are based on exploiting the self-correlations
structure of a time series. Box jenkins models are also known by the name of
ARIMA models, which is an acronym for integrated auto-regressive mobile av-
erage. The self-correlation function plays a central role in the development of
these models, this is the characteristic that distinguishes the ARIMA model
from the other methods mentioned above. As all these forecasts are handled
through models, we denote the time series, denoting the time series of interest
as D1 , D2 , ...Dn . We are going to assume initially that the series is stationary.
In this way E = Di = µ and var = Di = σ 2 ∀i ∈ 1, 2, ....n. Practically speaking,
in the seasonality there is no growth or decrease in the series, and the variation
remains relatively constant. This is important to denote that seasonality does
not imply independence. Therefore it is possible to evaluate Di and Dj they
are dependent random variables when i 6= j although their marginal density
functions are the same.
98 J. Romero-Gelvez et al.

The assumption of seasonality implies that the marginal distribution of two

observations separated by the same time interval are the same. This means that
Dt and Dt+1 they have the same distribution as Dt+m y Dt+m+1 for any m ≥ 1.
This implies that the covariance of Dt and Dt+1 is exactly the same covari-
ance of Dt+m and Dt+m+1 . Therefore , the covariance of these two observations
depends only on the number of periods that separate them. In this context, co-
variance is also known as self-covariance , we are comparing two values of the
same series separated by a fixed delay .Leave Cov(Dt+m , Dt+m+k) be the covari-
ance of Dt+m and Dt+m+k given by Cov(Dt+m , Dt+m+k) = E(Dt+m Dt+m+k ) −
E(Dt+m )E(Dt+m+k )∀k, int ≥ 1.
There are four main steps needed to build BOX-JENKINS forecast models

1. Data transformations: The BOX-JENKINS method is based on the start of a

series of stationary time. To be sure that the time series is stationary, several
steps are needed preliminarily. We know that differentiation eliminates trend
and seasonality. However, if the average of the series is relatively fixed, this
may be the case in which the variance is not constant, so a transformation
of the data is going to be required.
2. Identification of the model: This step refers exactly to which is the most
appropriate ARIMA model. The identification of the type of model is both
art and science, it is difficult if not impossible, to identify the model only the
series is examined. It is much more effective to study the self-correlations of
samples and partial self-correlations for the identification of patterns that
coincide with those of the known processes. In some cases, the self-correlation
structure will point to a simple AR or MA process, but it is more common
to mix these two terms to have a better fit.
3. Parameter estimation: Once the appropriate model has been identified, the
optimal values of the parameters of the model (i.e., a0 , a1 , . . . ., ap yb0 , b1 , . . . .,
bq ) must be determined, usually this step is carried out by means of least
squares adjustment methods or by the maximum likelihood method, this
step is performed by a computer program.
4. Forecasts: Once the model has been identified and the values of the optimal
parameters determined, the model provides forecasts of future values of the
series. The BOX-JENKINS models are more effective in providing one-step
forecasts, but can also provide multi- step forecasts.
5. Evaluation: The waste pattern (forecast errors) can give useful information
to see the quality of the model. The residuals must form a white noise ( ie,
random) process with zero means. When there are patterns in the waste, it
is suggested that the model can improve.

2.3 Long short-term memory networks

The bright idea of introducing self-loops to produce paths where the gradient
can flow for long durations is a core contribution of the first long short-term
memory (LSTM) model [11]. In this case, we mean that even for an LSTM with
 Demand Forecasting and Material Requirement Planning Optimization 99

Fig. 3. Block diagram of the LSTM recurrent network “cell.”

fixed parameters, the time scale of integration can change based on the input
sequence, because of the time constants.
The LSTM has been found hugely successful in many applications, such as
unconstrained handwriting recognition [8], speech recognition [7, 9] handwriting
generation [6], machine translation (Sutskever et al., 2014), image captioning
[14, 28, 30], and parsing [27] .
The LSTM block diagram is illustrated in Fig. 3. The corresponding for-
ward propagation equations are given below, for a shallow recurrent network
architecture.
Cells are connected recurrently to each other, replacing the usual hidden
units of ordinary recurrent networks. An input feature is computed with a reg-
ular artificial neuron unit. Its value can be accumulated into the state if the
sigmoidal input gate allows it. The state unit has a linear self-loop whose weight
is controlled by the forget gate. The output of the cell can be shut off by the
output gate. All the gating units have a sigmoid nonlinearity, while the input
unit can have any squashing nonlinearity. The state unit can also be used as an
extra input to the gating units. The black square indicates a delay of a single
time step.  
(t) f (t) f (t−1) 
X X
fi = σ bfi + Ui,j xj + Wi,j hj , (1)
j j

 
(t) (t) (t−1) (t−1)  (t) (t−1) 
X X
si = fi si + gi σ bi + Ui,j xj + Wi,j hj , (2)
j j

 
(t) g (t) g (t−1) 
X X
gi = σ bgi + Ui,j xj + Wi,j hj . (3)
j j
100 J. Romero-Gelvez et al.

(t)
(t)
hi = tanh sti qi , (4)

 
(t) 0 (t) 0 (t−1) 
X X
qi = σ b0i + Ui,j xj + Wi,j hj , (5)
j j

2.4 The prophet forecasting model

We use Prophet open-source software in Python [26] based in a decomposable

time series model [10] components: trend, seasonality, and holidays. They are
combined in the following equation:

y(t) = g(t) + s(t) + h(t) + t (6)

Were g(t) is the trend function which models non periodic changes in the
value of the time series, s(t) represents periodic changes, and h(t) represents
the effects of holidays which occur on potentially irregular schedules over one
or more days. The error term t represents any idiosyncratic changes that are
not accommodated by the model; later we will make the parametric assumption
that t is normally distributed.

3 Solution method

Open source tools in optimization are increasing popularity over the last years
and can be observed in works like [24] for the optimization process in supply
chain management and [19] in machine learning applications. We describe our
approach for using open source tools in MRP with dynamic demand as follows.

– Forecasting Model selection: In order to use the method that generates

less error t . First, we apply the SARIMA model through the ASTSA library
1
, from the R programming language. In addition, Fb-prophet 2 and Lstm-
Keras 3 are also used with the Python programming language. Next, we
select the method with the least error of them.
– JuMP-Julia for Mathematical Programming4 : JuMP it is a package
for mathematical optimization and operations research.
– Cbc Solver5 : Cbc it is a free solver, that let us solve linear and integer
optimization problems.
– IDE: IPython/IR/IJulia/Jupyter notebooks and Google-Colab.
1
https://cran.r-project.org/web/packages/astsa/index.html
2
https://facebook.github.io/prophet/docs/quickstart.html
3
https://keras.io/
4
http://www.juliaopt.org/JuMP.jl/v0.13/
5
https://github.com/coin-or/Cbc
 Demand Forecasting and Material Requirement Planning Optimization 101

Fig. 4. Methodological approach

In this paper, we propose a methodology based on a combination of Forecast-

ing and MIP techniques to help the managers of the fragrance industry to decide
the best inventory and supply policies. The steps followed in the methodology
are shown in the Fig. 4.
The methodology is applied to a particular manufacturing case study: to
help the managers of the fragrance industry to achieve their goals related to
minimizing inventory and minimizing significant deviations in their forecasts.
First, the forecasting job is performed by using the box-Jenkins methodology al-
though ASTSA package in R, the Fb-Prophet package and Lstmn with KERAS
in Python, next, the Master Production Schedule, is created based on the pre-
vious forecasts, and the MRP optimization is performed to get the final MRP
plan. The results of the MRP model are evaluated comparing their results with
the real demand and improved by adjusting the historical data demand records.

4 Results

We apply the solution method to a company dedicated to the creation develop-

ment, manufacture, and marketing of flavors and fragrances for the Colombian
industry. A step by step application is described as follows:

4.1 Forecasting the principal SKU

First, the historical demand data is plotted in order to isolate the data patterns.
Later we apply the SARIMA model by using the R language Fig. 5, also the
Fb-Prophet package and Lstmn using KERAS and Google Colab Fig. 6.
Table 1 shows the difference in forecasts generated and their associated errors
in the last row. The techniques with the lowest associated error are SARIMA
and LSTMN, so the predicted demand generated by them will be used, which
can be seen in Table 1.
As it can be seen in Table 2 the future demand for 9 months is the principal
input for the MRP formulation, and it will be represented on the subset Dj .
102 J. Romero-Gelvez et al.

(a) Historical demand data plot (b) Forecasting for 2019 (monthly).

Fig. 5. Forecast using SARIMA in R language

(a) Fb-Forecast data plot (b) Lstm-Keras

Fig. 6. Forecast using SARIMA in R language

Table 1. Comparison of techniques and RMSE

Technique SARIMA FB-PROPHET LSTMN

m1 122,1863 220,064847 222,8
m2 144,89 293,060167 185,6
m3 114,5067 197,096387 156
m4 108,1722 71,020883 99,6
m5 108,3905 182,760315 164
m6 112,6212 20,875617 240
m7 119,4588 186,352169 141,6
m8 121,4177 419,670517 114
m9 123,3766 132,994213 92
RMSE 39,95761 127,764023 53,53296
 Demand Forecasting and Material Requirement Planning Optimization 103

Table 2. Forecasted Demand Dj

Apr May Jun Jul Aug

88.34232 98.20327 104.55308 108.39207 110.74301
Sept Oct Nov Dec
112.17886 113.05630 113.59244 113.92004

4.2 Using the optimization model for MRP

In this item, we use the Julia programing language and the package JuMP for
mathematical optimization, the code for the problem it can be seen as follows.

using JuMP,Cbc,NamedArrays,DataFrames
filas=size(D,1)
col=size(D,2)
mrp=Model(solver=CbcSolver())
@variables mrp begin
x[1:filas,1:col]>=0
d[1:filas,1:col]>=0
end
T=col
@objective(mrp,Min,sum( x[i,j]*((T-j)) for i=1:filas, j=1:col))
for i=1:filas,t=1:col
@constraint(mrp, sum(x[i,s] for s=1:t-LT[i,1])+I[i,1]>=sum(D[i,s]
+sum(R[i,j]*x[j,s] for j=1:filas) for s=1:t))
end
@constraint(mrp,x-d.*LS.>=0)
@constraint(mrp,d-x/1000000000000000.>=0)
status=solve(mrp)
println(getobjectivevalue(mrp))
println(DataFrame(getvalue(x)))
println(DataFrame(getvalue(d)))

We use the Ri,j bill of manufacture that shows in the Table 3 as follows:

Table 3: Bill of manufacture Ri,j

j1 j2 j3 j4 j5 . . . j34
i1 0 0 0 0 0 ... 0
i2 253.816 0 0 0 0 ... 0
i3 0.4 0 0 0 0 ... 0
i4 20 0 0 0 0 ... 0
i5 20 0 0 0 0 ... 0
i6 105.784 0 0 0 0 ... 0
i7 0.1 0 0 0 0 ... 0
104 J. Romero-Gelvez et al.

Table 3 continued from previous page

j1 j2 j3 j4 j5 . . . j34
i8 0 0 0 0 6.44 ... 0
i9 0 0 0 0 0.04 ... 0
i10 0 0 0 0 0.24 ... 0
i11 0 0 0 0 0.2 ... 0
i12 0 0 0 0 2.312 ... 0
i13 0 0 0 0 1.4 ... 0
i14 0 0 0 0 1.16 ... 0
i15 0 0 0 0 1.04 ... 0
i16 0 0 0 0 0.88 ... 0
i17 0 0 0 0 0.85 ... 0
i18 0 0 0 0 0.7 ... 0
i19 0 0 0 0 0.48 ... 0
i20 0 0 0 0 0.456 ... 0
i21 0 0 0 0 0.4 ... 0
i22 0 0 0 0 0.376 ... 0
i23 0 0 0 0 0.296 ... 0
i24 0 0 0 0 0.2 ... 0
i25 0 0 0 0 0.124 ... 0
i26 0 0 0 0 0.08 ... 0
i27 0 0 0 0 0.098 ... 0
i28 0 0 0 0 0.044 ... 0
i29 0 0 0 0 0.04 ... 0
i30 0 0 0 0 0.04 ... 0
i31 0 0 0 0 0.04 ... 0
i32 0 0 0 0 0.04 ... 0
i33 0 0 0 0 0.024 ... 0
i34 0 0 0 0 2 ... 0

4.3 Final Results

Table 4: Final results for every sku i and their needs over time in
kgs

t1 t2 t3 t4 t5 t6 t7 t8
sku1 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
sku2 0.0 0.0 0.0 0.0 22422.7 24925.6 26537.2 27511.6
sku3 0.0 0.0 0.0 0.0 35.3 39.3 41.8 43.4
sku4 0.0 0.0 0.0 0.0 1766.8 1964.1 2091.1 2167.8
sku5 0.0 0.0 0.0 0.0 1766.8 1964.1 2091.1 2167.8
sku6 0.0 0.0 0.0 0.0 9345.2 10388.3 11060.0 11466.1
sku7 0.0 0.0 0.0 0.0 8.8 9.8 10.5 10.8
 Demand Forecasting and Material Requirement Planning Optimization 105

Table 4 continued from previous page

t1 t2 t3 t4 t5 t6 t7 t8
sku8 11378.5 12648.6 13466.4 13960.9 14263.7 14448.6 14561.7 14630.7
sku9 70.7 78.6 83.6 86.7 88.6 89.7 90.4 90.9
sku10 424.0 471.4 501.9 520.3 531.6 538.5 542.7 545.2
sku11 353.4 392.8 418.2 433.6 443.0 448.7 452.2 454.4
sku12 4084.9 4540.9 4834.5 5012.0 5120.8 5187.2 5227.7 5252.5
sku13 2473.6 2749.7 2927.5 3035.0 3100.8 3141.0 3165.6 3180.6
sku14 2049.5 2278.3 2425.6 2514.7 2569.2 2602.5 2622.9 2635.3
sku15 1837.5 2042.6 2174.7 2254.6 2303.5 2333.3 2351.6 2362.7
sku16 1554.8 1728.4 1840.1 1907.7 1949.1 1974.3 1989.8 1999.2
sku17 1501.8 1669.5 1777.4 1842.7 1882.6 1907.0 1922.0 1931.1
sku18 1236.8 1374.8 1463.7 1517.5 1550.4 1570.5 1582.8 1590.3
sku19 848.1 942.8 1003.7 1040.6 1063.1 1076.9 1085.3 1090.5
sku20 805.7 895.6 953.5 988.5 1010.0 1023.1 1031.1 1036.0
sku21 706.7 785.6 836.4 867.1 885.9 897.4 904.5 908.7
sku22 664.3 738.5 786.2 815.1 832.8 843.6 850.2 854.2
sku23 523.0 581.4 619.0 641.7 655.6 664.1 669.3 672.5
sku24 353.4 392.8 418.2 433.6 443.0 448.7 452.2 454.4
sku25 219.1 243.5 259.3 268.8 274.6 278.2 280.4 281.7
sku26 141.3 157.1 167.3 173.4 177.2 179.5 180.9 181.7
sku27 173.2 192.5 204.9 212.4 217.1 219.9 221.6 222.6
sku28 77.7 86.4 92.0 95.4 97.5 98.7 99.5 100.0
sku29 70.7 78.6 83.6 86.7 88.6 89.7 90.4 90.9
sku30 70.7 78.6 83.6 86.7 88.6 89.7 90.4 90.9
sku31 70.7 78.6 83.6 86.7 88.6 89.7 90.4 90.9
sku32 70.7 78.6 83.6 86.7 88.6 89.7 90.4 90.9
sku33 42.4 47.1 50.2 52.0 53.2 53.8 54.3 54.5
sku34 3533.7 3928.1 4182.1 4335.7 4429.7 4487.2 4522.3 4543.7

5 Discussion and future investigation

Future investigation must be related to DDMRP demand driven MRP and ma-
chine learning techniques related to forecasting methods. The first affirmations
is related with methods for maintain a minimum and maximum intervals for in-
ventory management over every sku. The second affirmation is related to method
that helps us to obtain better and more accurates forecasts for future demand.

6 Conclusions
The use of Open source optimization and forecasting tools allows to solve opti-
mization and predictive problems efficiently.
The forecasting problem is the principal problem in an MRP based company
and can be solved at scale using open source tools.
106 J. Romero-Gelvez et al.

The biggest contribution of this work is to show how to solve a real problem
over forecasting and optimization using three programming languages in order
to reduce error and improve the time scheduling for every stock keeping unit.
The company uses this implementation in order to plan his manufacturing
process at a larger scale including every product of their portfolio.

References
1. Altendorfer, K., Felberbauer, T., Jodlbauer, H.: Effects of forecast errors on optimal
utilisation in aggregate production planning with stochastic customer demand.
International Journal of Production Research 54(12), 3718–3735 (2016)
2. Bousqaoui, H., Achchab, S., Tikito, K.: Machine learning applications in supply
chains: long short-term memory for demand forecasting. In: International Confer-
ence of Cloud Computing Technologies and Applications. pp. 301–317. Springer
(2017)
3. Box, G.E., Jenkins, G.M., Reinsel, G.C., Ljung, G.M.: Time series analysis: fore-
casting and control. John Wiley & Sons (2015)
4. Cox, J.F., Blackstone, J.H.: APICS dictionary. Amer Production & Inventory
(2002)
5. Disney, S.M., Farasyn, I., Lambrecht, M., Towill, D.R., Van de Velde, W.: Tam-
ing the bullwhip effect whilst watching customer service in a single supply chain
echelon. European Journal of Operational Research 173(1), 151–172 (2006)
6. Graves, A.: Generating sequences with recurrent neural networks. arXiv preprint
arXiv:1308.0850 (2013)
7. Graves, A., Jaitly, N.: Towards end-to-end speech recognition with recurrent neural
networks. In: International conference on machine learning. pp. 1764–1772 (2014)
8. Graves, A., Liwicki, M., Fernández, S., Bertolami, R., Bunke, H., Schmidhuber,
J.: A novel connectionist system for unconstrained handwriting recognition. IEEE
transactions on pattern analysis and machine intelligence 31(5), 855–868 (2008)
9. Graves, A., Mohamed, A.r., Hinton, G.: Speech recognition with deep recurrent
neural networks. In: 2013 IEEE international conference on acoustics, speech and
signal processing. pp. 6645–6649. IEEE (2013)
10. Harvey, A.C., Peters, S.: Estimation procedures for structural time series models.
Journal of Forecasting 9(2), 89–108 (1990)
11. Hochreiter, S., Schmidhuber, J.: Long short-term memory. Neural computation
9(8), 1735–1780 (1997)
12. Hopp, W.J., Spearman, M.L.: Factory physics. Waveland Press (2011)
13. Jacobs, F.R.: Manufacturing planning and control for supply chain management.
McGraw-Hill (2011)
14. Kiros, R., Salakhutdinov, R., Zemel, R.S.: Unifying visual-semantic embeddings
with multimodal neural language models. arXiv preprint arXiv:1411.2539 (2014)
15. Lee, T., Adam Jr, E.E.: Forecasting error evaluation in material requirements plan-
ning (mrp) production-inventory systems. Management Science 32(9), 1186–1205
(1986)
16. Lee, T., Shih, W.: Optimal forecast biasing in theoretical inventory models. THE
INTERNATIONAL JOURNAL OF PRODUCTION RESEARCH 27(5), 809–830
(1989)
17. Li, Q., Disney, S.M.: Revisiting rescheduling: Mrp nervousness and the bullwhip
effect. International Journal of Production Research 55(7), 1992–2012 (2017)
 Demand Forecasting and Material Requirement Planning Optimization 107

18. Mabert, V.A.: The early road to material requirements planning. Journal of Op-
erations Management 25(2), 346–356 (2007)
19. Moreno, R.H., Garcia, O., et al.: Model of neural networks for fertilizer recom-
mendation and amendments in pasture crops. In: 2018 ICAI Workshops (ICAIW).
pp. 1–5. IEEE (2018)
20. Nahmias, S., Cheng, Y.: Production and operations analysis, vol. 6. McGraw-hill
New York (2005)
21. Orlicki, J.A.: Material requirements planning: the new way of life in production
and inventory management. McGraw-Hill (1975)
22. Priore, P., Ponte, B., Rosillo, R., de la Fuente, D.: Applying machine learning
to the dynamic selection of replenishment policies in fast-changing supply chain
environments. International Journal of Production Research pp. 1–15 (2018)
23. Ptak, C.A., Smith, C.: Demand Driven Material Requirements Planning
(DDMRP). Industrial Press, Incorporated (2016)
24. Romero-Gelvez, J.I., Gonzales-Cogua, W.C., Herrera-Cuartas, J.A.: Cvrptw model
for cargo collection with heterogeneous capacity-fleet. In: International Conference
on Applied Informatics. Springer (2019)
25. Sridharan, V., BERRY, W.L.: Master production scheduling make-to-stock prod-
ucts: a framework for analysis. The International Journal of Production Research
28(3), 541–558 (1990)
26. Taylor, S.J., Letham, B.: Forecasting at scale. The American Statistician 72(1),
37–45 (2018)
27. Vinyals, O., Kaiser, Ł., Koo, T., Petrov, S., Sutskever, I., Hinton, G.: Grammar
as a foreign language. In: Advances in neural information processing systems. pp.
2773–2781 (2015)
28. Vinyals, O., Toshev, A., Bengio, S., Erhan, D.: Show and tell: A neural image
caption generator. In: Proceedings of the IEEE conference on computer vision and
pattern recognition. pp. 3156–3164 (2015)
29. Voß, S., Woodruff, D.L.: Introduction to computational optimization models for
production planning in a supply chain, vol. 240. Springer Science & Business Media
(2006)
30. Xu, K., Ba, J., Kiros, R., Cho, K., Courville, A., Salakhutdinov, R., Zemel, R.,
Bengio, Y.: Show, attend and tell: Neural image caption generation with visual
attention. arXiv preprint arXiv:1502.03044 (2015)